Research map / Vision-language and multimodal / Cross-modal learning
DASC: Robust Dense Descriptor for Multi-Modal and Multi-Spectral Correspondence Estimation
Seungryong Kim, Dongbo Min, Bumsub Ham, Minh N. Do, Kwanghoon Sohn
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add adaptive self-correlation in place of SSD in LSS (ablation) | LSS (no max-pool) · Middlebury (illumination/exposure) | bad-pixel error | Improved |
| add randomized pooling (random center pixel) in place of center-biased pooling (ablation) | DASC · Middlebury (illumination/exposure) | bad-pixel error | Improved |
| add learned sampling pattern (LRP) on randomized pooling (ablation) | DASC+RP · Multi-modal/multi-spectral (WTA average) | error rate: 12.18 -> 11.80 | Improved |
| add learned sampling pattern (LRP) on randomized pooling (ablation) | DASC+RP · Multi-modal/multi-spectral (SIFT flow average) | error rate: 9.68 -> 7.64 | Improved |
| use efficient constant-time edge-aware filter computation instead of brute force (ablation) | DASC · Runtime 463x370 | time (s): 128s -> 1.3s | Improved |
| use efficient computation instead of brute force (ablation) | DASC · Runtime 800x600 | time (s): 256s -> 2.1s | Improved |
| use asymmetric instead of symmetric adaptive self-correlation measure (ablation) | DASC · Middlebury (illumination/exposure) | bad-pixel error | No clear effect |
| use edge-aware filter weights instead of box or Gaussian filtering (ablation) | DASC · Middlebury (illumination/exposure) | bad-pixel error | Improved |
| increase patch size N (ablation) | DASC · Middlebury (illumination/exposure) | bad-pixel error | Hurt |
| increase descriptor dimension L (ablation) | DASC · Middlebury (illumination/exposure) | bad-pixel error | Improved |
| increase support window size M (saturates near 31x31) (ablation) | DASC · Middlebury (illumination/exposure) | bad-pixel error | Improved |
| increase number of log-polar circular points Nc (ablation) | DASC · Middlebury (illumination/exposure) | bad-pixel error | No clear effect |
Closest papers by meaning
- Dense Cross-Modal Correspondence Estimation With the Deep Self-Correlation Descriptor (TPAMI 2020 · similarity 0.91)
- Dense Correspondences across Scenes and Scales (TPAMI 2016 · similarity 0.87)
- Deep Self-correlation Descriptor for Dense Cross-Modal Correspondence (ECCV 2016 · similarity 0.86)
- Efficient LoFTR: Semi-Dense Local Feature Matching with Sparse-Like Speed (CVPR 2024 · similarity 0.83)
- Self-Supervised Spatial Correspondence Across Modalities (CVPR 2025 · similarity 0.83)
- UFM: A Simple Path towards Unified Dense Correspondence with Flow (NeurIPS 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
